I use sorted sets for a lot of stuff (timelines mostly). A few other examples are job queuing, error logs (I write all error messages to a fixed length list), and rate limiting. Actually that last one works pretty well. The key is something like:

  v1:limiter:UNIQUEID:TYPE:HHMM

where UNIQUEID is a username or IP address, etc, TYPE is the specific action to limit (login, signup, etc), and HHMM is a quantized timestamp (e.g. 1205, 1210). You increment this key every time the action occurs and use an expire to keep it clean. It's not perfect, but it does the trick for most usecases.

I also make pretty judicious use of databases to group different classes of data together. Primarily into 3 groups actually: transient data, persistent/customer data, and internal stuff like the error logs and job queues. That way I can dump data at different intervals for each group (I use a tool I wrote called redis-dump: https://github.com/delano/redis-dump).